On 09/03/2017 09:57 AM, scott wrote:
>> I guess POV-Ray is pretty much a dying technology at this point. I have
>> no idea what the cool kids are using these days; probably some
>> GPU-accelerated polygon renderer with global light transport.
>
> I used to use POV a lot for rendering CAD models that I had generated
> elsewhere. Now we bought software called "KeyShot" at work, the ease and
> speed that you can create photorealistic renders and animations is
> insane.

Yeah, I suspect anything GPU-powered is going to out-strip POV-Ray by 
many orders of magnitude. Indeed, there are WebGL demos that do stuff 
*in a web browser* that would take months to render in POV-Ray! (Trouble 
is, as I discovered, rendering an entire scene as a single monolithic 
shader doesn't scale beyond a few objects.)

>> It's a shame there isn't something modern that has a scene description
>> language like the SDL. But then again, if you actually have the talent
>> to model stuff, what do you need SDL for?
>
> And if you have the talent to program, why use SDL? Recently I've been
> writing a C#/OpenGL Mandelbulb animation renderer.

I have literally no idea how you would even *begin* to do something like 
that.

I used to assume that a 3D card takes a model and then renders it. Then 
I started reading some of the OpenGL spec, and realised that actually, 
the 3D card does *almost nothing*. It merely knows how to draw 
texture-mapped polygons really, *really* fast. If you want shadows, 
reflection, refraction, scattering... tough. You can't have it. (Or 
rather, you *can*... if you spend many, many months programming it all.)

The OpenGL API is huge, complex, and mostly undocumented. It's also 
extremely imperative. And you can't debug it. (I did see a while back a 
rather amusing page listing 35 different ways to accidentally render a 
completely black image...)

That said, I do wonder if there's a way to do physically-correct depth 
of field rendering using only polygon graphics. Like, if you could 
render multiple images from different angles and sum them... It would be 
far too slow for realtime, but I think you could do it.

Post a reply to this message